Wednesday, May 18, 2016 4:00pm - 5:00pm
Watson CIT - SWIG Boardroom (CIT241)
Inference of Genetic and Signal Networks from Molecular Measures
The dependence of variables in a complex dynamical system or stochastic process can be represented by a direct graph. We present techniques to infer these graphs for models of Genetic Networks or Protein Interactions from time series of measures, respectively, of gene expressions and protein concentrations. Gene Networks are modeled by a particular Markovian process, with several constraints, called Probabilistic Genetic Network (PGN). To illustrate the versatility of this model, we present an input-output model of the cell cycle control, implemented by a PGN. After, a technique for topology inference of PGN, based on the mean condition entropy estimation, is presented. This technique is used to estimate the genetic network of the Plasmodium Falciparum malaria parasite and to make annotations of these genes. This inference technique is improved considering the distribution of connections per gene in the network. Additionally, the question of integrating in the model other measures, besides expression, is discussed. The inference of protein interactions in Signal Networks is modeled by Reaction Kinetics equations. The correction of a proposed model is validated comparing the concentrations predicted and the ones measured experimentally. In the network search, a family of networks is generated automatically. This approach has been applied for studying the Ras-MAPK network. Some open problems are discussed in this context such as isolation of the subsystem studied.
Hosted by Pedro Felzenszwalb, and Basilis Gidas